[gmx-users] MM dihedral scanning

[Justin Lemkul]

On 1/12/17 10:21 PM, Mohsen Ramezanpour wrote:
> On Thu, Jan 12, 2017 at 6:31 PM, Justin Lemkul <jalemkul at vt.edu> wrote:
>
>>
>>
>> On 1/12/17 7:47 PM, Mohsen Ramezanpour wrote:
>>
>>> On Thu, Jan 12, 2017 at 4:24 PM, Justin Lemkul <jalemkul at vt.edu> wrote:
>>>
>>>
>>>>
>>>> On 1/12/17 3:18 PM, Mohsen Ramezanpour wrote:
>>>>
>>>> Dear Gromacs users,
>>>>>
>>>>> For parameterization of a molecule in Charmm36, I have got the QM
>>>>> scanning
>>>>> and partial charges from GAMMP server. However, the fitted parameters
>>>>> are
>>>>> not good enough.
>>>>>
>>>>>
>>>> That's very surprising.  What's wrong with what GAAMP gave you?
>>>>
>>>> The dihedral has two local minima in both QM and fitted ones both from
>>> GAAMP. The angle for the minima are okay but the corresponding depths are
>>> not.
>>> In fact, the depth for the first local minimum is larger than second one
>>> in
>>> QM, while the situation is reverse in MM profile with fitted parameters.
>>> This makes the dihedral to be more (statistically) in wrong angle (in
>>> local
>>> minimum which is not the most favourable one).
>>> GAAMP, unfortunately, did not work well with my case (some critical
>>> partial
>>> charges and critical dihedrals).
>>>
>>>
>>> I decided to do the MM scanning and try to get better parameters for the
>>>>
>>>>> dihedral.
>>>>>
>>>>> Unfortunately, I do not have any experience with this part, and I could
>>>>> not
>>>>> find any tutorial for how to do this in Gromacs.
>>>>> I was wondering if you are aware of any tutorial which could help me to
>>>>> overcome this challenging step.
>>>>>
>>>>>
>>>>> Tutorial (CGenFF theory is the same as CHARMM, by design):
>>>> http://mackerell.umaryland.edu/~kenno/cgenff/download.php#tutor
>>>>
>>>> Fitting program and other resources:
>>>> http://mackerell.umaryland.edu/~kenno/lsfitpar/
>>>> http://mackerell.umaryland.edu/ff_dev.shtml
>>>>
>>>> Obviously, these are all CHARMM-centric approaches and frankly the
>>>> modules
>>>> within CHARMM make parametrization rather straightforward (not "easy,"
>>>> mind
>>>> you, but straightforward).  Since I began working with CHARMM, it has
>>>> become indispensable in my daily routine.
>>>>
>>>> If you want to do things in GROMACS, the main issue is that you will have
>>>> to do MM scans in a more manual fashion, by restraining the target
>>>> dihedrals (very strongly) in a series of configurations (typically at
>>>> intervals of 15 degrees over a full rotation) while allowing the rest of
>>>> the molecule to relax to match the QM.
>>>>
>>>
>>> If I got it right, I must do EM on the molecule while only the desired
>>> dihedral is fixed in a specific angle. Which aspect should match with QM?
>>> If you mean structurally, what is the criteria for matching (RMSD?.)?
>>>
>>>
>> "Match" in this case means "treat equivalently," therefore only one
>> constraint (restraint in the MM) should be applied while allowing the rest
>> of the molecule to relax freely.  You do have a difficult case because each
>> of your molecules is symmetric; this means the same dihedral term is
>> affecting multiple torsions.
>>
> So, probably I should 4 dihedrals and try to optimize all at once?!(
> because all 4 dihedrals of O-C-CH2-CH3 seems equivalent to me). Am I right?
>

No, there are 2 O-C-CH2-CH3 dihedrals, not 4.

>>
>>
>>>
>>>   Deactivate the restraint, obtain the potential energy of the molecule
>>> via
>>>
>>>> mdrun -rerun and plot as a function of the dihedral.
>>>>
>>>
>>> This should be a zero step EM, right? The molecule should not be allowed
>>> to
>>> change its conformation.
>>>
>>>
>> No, a zero-step MD.  EM actually changes the coordinates before step zero.
>>
>>   A bit of shell scripting and careful topology modification and this can
>>>
>>>> be done.
>>>>
>>>>
>>> Two more questions on this part:
>>> 1) I am using the QM scanning data and partial charges from GAAMP. When I
>>> do this MM scanning, do I need to exclude any 1-4 interactions or I can
>>> behave this dihedral as other dihedrals?
>>>
>>
>> 1-4 interactions are always at full strength in CHARMM.
>>
>> 2) this dihedral is part of a lipid.
>>> Do I need to do these on only "one Lipid in vacuum"  or
>>> OR
>>> on all lipids in "a bilayer in vacuum" or in a "bilayer in solvent"
>>>
>>> I think it should be  "one Lipid in vacuum".
>>>
>>>
>> One model compound in vacuum, from which you will construct the lipid.
>>
> How if this compound (which is small part of the lipid) is charged? Should
> it be still in vacuum?

Yes.  The fact that a molecule is irrelevant except for when considering the 
dipole moment.

> Is there any specific consideration to be made in zero-step MD? e.g. a big
> simulation box or specific parameter in .mdp file?
>

Infinite cutoffs, no PBC.

-Justin

-- 
==================================================

Justin A. Lemkul, Ph.D.
Ruth L. Kirschstein NRSA Postdoctoral Fellow

Department of Pharmaceutical Sciences
School of Pharmacy
Health Sciences Facility II, Room 629
University of Maryland, Baltimore
20 Penn St.
Baltimore, MD 21201

jalemkul at outerbanks.umaryland.edu | (410) 706-7441
http://mackerell.umaryland.edu/~jalemkul

==================================================